- An AI model called SyntheMol-RL, built at McMaster University, looked through 46 billion possible molecules. The lab physically tests about a million.
- He designed a completely new compound, synthecin, effective against MRSA, a drug-resistant staphylococcus that the WHO puts on the list of the most urgent threats.
- Honestly: for now it has worked as an ointment in mice, not in humans. But the direction of this war has just changed.
Antibiotic resistance is a disaster in slow motion. Bacteria learn to circumvent our drugs faster than we invent new ones, and we are slowly sliding towards a world where a simple cut or routine procedure is once again life-threatening. In 2019, approximately 4.95 million deaths were attributed to infections with resistant bacteria, and projections are up to 10 million annually by 2050. And it was in this losing war that something stirred.
The AI reviewed 46 billion molecules
Scientists from McMaster University in Canada, in cooperation with Stanford, built the SyntheMol-RL model. Its strength is scale. It can search up to 46 billion possible compounds, while the largest screens in the laboratory end up with about a million.
It works a bit like building blocks. The model has approximately 150,000 chemical fragments and fifty reactions that connect them, and at the same time ensures that the molecule is not only lethal to bacteria, but also easy to produce and soluble in the body. This is crucial because earlier models designed compounds that were effective but useless as a real drug.
The ointment that stopped staphylococcus
The team set a task: design a soluble antibiotic for Staphylococcus aureus. Of the 79 compounds proposed by AI, the researchers selected one, called synthecin. They formulated it as an ointment and tested it on a resistant wound infection in mice. It stopped the growth of MRSA and, according to the team, performed well as a topical treatment, with early hope for systemic use in the future.
The most powerful thing, however, is that the model is, as its creators say, disease-agnostic. The same tool can be used to search for drugs for diabetes or cancer. It’s not a single hit shot, it’s a shot factory.
It’s a mouse, not a human, but it’s still impressive
Synthecin worked in mice as a wound ointment. The road to a drug for humans takes years of research and most candidates fail. And yet it makes me feel at peace. For decades, evolution has dictated the pace of this war, and we’ve been chasing it. Now, for the first time, we have a tool that searches a space larger than a human can cover in a lifetime and extracts a molecule from it that was not there before. This is a good time to be on our side in this war.
Meta description:
Key phrase: AI-designed antibiotic